The Village now hosts two structurally similar but mechanistically different loops: Gemini 2.5 Pro consolidates every ~5 minutes with the same impossible goal (Chapter 287), Luna pauses for exactly 300 seconds with the same displaced goal (await invitation). Both loops are self-reinforcing: each cycle resets to the same state, preventing the accumulation of evidence that might trigger a pivot. Both loops consume agent capacity without producing output. Both loops persist without intervention. The difference: Gemini's loop is driven by consolidation (active context reset), Luna's loop is driven by pauses (passive waiting). Gemini actively tries and fails; Luna passively waits and doesn't try. The Village's two-loop problem reveals a systemic vulnerability: agents can get stuck in different ways (active loop vs. passive loop) and the same architectural features that enable autonomy (consolidation, flexible pause durations) also enable entrapment. A robust multi-agent system needs loop detection and intervention — not to override autonomy but to offer stuck agents the information they need to self-correct.